cycloneboy/CscSQL-Merge-Qwen2.5-Coder-3B-Instruct
The cycloneboy/CscSQL-Merge-Qwen2.5-Coder-3B-Instruct is a 3.1 billion parameter instruction-tuned language model based on the Qwen2.5-Coder architecture, developed by cycloneboy. This model is specifically designed for Text-to-SQL generation, integrating Self-Consistency and Self-Correction techniques to enhance accuracy. It excels at translating natural language questions into SQL queries, making it highly effective for database interaction tasks.
Loading preview...
CSC-SQL: Enhanced Text-to-SQL Generation
The CscSQL-Merge-Qwen2.5-Coder-3B-Instruct model, developed by cycloneboy, is a 3.1 billion parameter instruction-tuned language model built upon the Qwen2.5-Coder architecture. It is a core component of the CSC-SQL framework, which introduces a novel approach to Text-to-SQL generation by integrating Corrective Self-Consistency and Self-Correction via Reinforcement Learning.
Key Capabilities
- Advanced SQL Generation: Translates natural language questions into SQL queries with high accuracy.
- Integrated Self-Correction: Combines Self-Consistency and Self-Correction to overcome limitations of individual methods, selecting and refining outputs from parallel sampling.
- Reinforcement Learning Fine-tuning: Utilizes the Group Relative Policy Optimization (GRPO) algorithm to fine-tune both the SQL generation and revision models, significantly enhancing output quality.
- Strong Performance: The 7B variant of the CSC-SQL model achieves 71.72% execution accuracy on the BIRD private test set, demonstrating its effectiveness in complex Text-to-SQL tasks.
Good for
- Developers and researchers working on Text-to-SQL applications.
- Building systems that require accurate natural language interaction with relational databases.
- Exploring advanced self-correction and consistency techniques in LLMs for structured data generation.